Collaborative Data Mining using Incentives and Multi-Party Secure Communications

Security and privacy play an important role in collaborative data mining when multiple competing parties are involved. The environment is expected to be a non-cooperative and there is no guarantee that all parties provide compatible or correct inputs. To encourage them to provide correct inputs, many techniques came into existence. One such method was recently proposed by Kantarcioglu et al. based on the incentives. Incentives let the competing parties to give meaningful inputs. In this paper we build mechanisms based on incentives and secure multi-party communications to ensure security and privacy in collaborative data mining. Our prototype application demonstrates the proof of concept. It takes data from multiple competing parties and verify the inputs for their correctness. Based on the correct inputs, the incentives of the party which gives input will be increased. When misbehavior is encountered, the incentives are reduced and the party is asked to provide correct inputs again. On giving the correct input, the reduced incentives are added again. This kind of technique and also a mechanism for secure multi-party communications makes the system useful for secure and privacy preserving collaborative data mining. The empirical results are encouraging.